Cookiy MCP — AI User Research Skills for Your Agent

One-command setup for Cookiy MCP server in your AI coding clients.

Cookiy gives your AI agent user-research skills — design interview guides, conduct AI-moderated interviews with real or simulated participants, and generate analysis reports — all through the Model Context Protocol.

Quick Start

npx cookiy-mcp

That's it. The CLI auto-detects your installed AI clients and configures them.

Supported Clients

| Client | Detection | Setup Method | |--------|-----------|-------------| | Claude Code | claude CLI | Automatic via CLI | | Cursor | ~/.cursor/ | JSON config | | VS Code (Copilot) | code CLI | JSON config | | Codex | codex CLI | Automatic via CLI | | Windsurf | ~/.windsurf/ | JSON config | | Cline | ~/.cline/ | JSON config | | OpenClaw | ~/.openclaw/ | Interactive OAuth + script |

Usage

# Default (production)
npx cookiy-mcp

# Specific environment
npx cookiy-mcp dev

# Only configure one client
npx cookiy-mcp --client cursor

# OpenClaw (interactive OAuth flow)
npx cookiy-mcp --client openclaw

# Preview without writing
npx cookiy-mcp --dry-run

# Remove configuration
npx cookiy-mcp --remove

# Skip confirmation
npx cookiy-mcp -y

What You Get — 5 Skills

Once connected, your AI agent gains these skill modules:

Study Creation

Describe your research goal in plain language, and Cookiy creates a complete study with an AI-generated discussion guide.

• run_study_quickstart — One-step study creation
• cookiy_study_create — Create a study asynchronously
• cookiy_media_upload — Upload images as study attachments

AI Interview

Simulate user interviews with AI personas — no real participants needed. Get preliminary insights in minutes.

• cookiy_fake_interview_generate — Queue AI-to-AI interview simulations
• cookiy_fake_interview_status — Check simulation job progress
• cookiy_interview_list — List interviews for a study
• cookiy_interview_playback_get — Get transcripts and recordings

Discussion Guide

Auto-generated interview scripts you can edit. Preview the impact of changes before applying.

• cookiy_guide_get — Retrieve the current guide
• cookiy_guide_patch — Apply changes with revision lock
• cookiy_guide_impact — Preview patch impact without saving
• cookiy_guide_status — Check guide generation status
• update_study_guide — Orchestrated get → impact → patch flow

Recruitment

Recruit real respondents through third-party panel providers to participate in AI-moderated interviews.

• cookiy_recruit_cint_create — Launch or reconfigure recruitment
• cookiy_recruit_status — Monitor recruitment progress

Report & Insights

Auto-generate analysis reports from completed interviews. Manage studies and track usage.

• cookiy_report_preview — Generate and share report links
• cookiy_study_get — Get study summary
• cookiy_study_list — List all studies
• cookiy_study_artifacts — Get study artifacts
• cookiy_balance_get — Check account balance

Example Workflow

After setup, ask your AI agent:

"Create a user research study about why users abandon shopping carts"

The agent will use Cookiy MCP skills to:

1. Create the study with AI-generated discussion guide
2. Run simulated AI interviews
3. Generate a research report with insights

Options

npx cookiy-mcp [server-url] [options]

Arguments:
  server-url              MCP server base URL, or environment alias:
                          prod, dev, dev2, preview, staging, test
                          (default: prod → https://s-api.cookiy.ai)

Options:
  --client <name>         Target specific client
                          (claudeCode, cursor, vscode, codex, windsurf, cline, openclaw)
  --name <server-name>    Override MCP server name (default: cookiy)
  --scope <scope>         Claude Code scope: user|project (default: user)
  --remove                Remove Cookiy MCP config
  --dry-run               Preview changes without writing
  -y, --yes               Skip confirmation
  -h, --help              Show help
  -v, --version           Show version

Requirements

• Node.js >= 18
• A Cookiy account (free to start)

How It Works

1. Validates the Cookiy MCP server via OAuth discovery endpoint
2. Detects which AI clients are installed on your machine
3. Writes the appropriate configuration for each client
4. Each client handles OAuth login on first use (except OpenClaw, which completes OAuth during setup)

Links

• Cookiy — AI-powered user research platform
• Cookiy Developer Portal — API keys and documentation

License

MIT